Author: Dann Corbit
Date: 11:36:08 09/03/99
Go up one level in this thread
On September 03, 1999 at 03:22:54, Terry Ripple wrote: >On September 02, 1999 at 20:31:41, Dann Corbit wrote: > >>An interesting thing about probability is that you might get 1000 heads in a row >>with a perfectly fair coin. In fact, that sequence is exactly as likely as any >>other run of 1000 trials (being 1/2^1000). >--------- > >Would you predict the same in human games as compared to what you are saying >about computers? Yes. Note that 1/2^1000 is a very small number. It's: 9.3326361850321887899e-302 Let's call that "really, really, really unlikely." Sometimes people think that computers are predictable. But here is something odd: For C.A.P., we calibrate the programs to use the exact same CPU time -- or as close as possible. We find that even the same program on equivalent hardware gives a slightly different answer about 1/4 of the time. It really does not matter why a different decision is made. For a computer, it might be that an IRQ got attention, or that the temperature changed -- altering the CPU frequency. Perhaps the program learned from before and so played somewhat differently. A large program may swap some blocks of the chess program to disk. A human might try to speculate and try something risky. He might have skipped breakfast and missed his morning coffee. For whatever reason, the physical world is full of random variables. We just can't get away from them. Ecc. 9:11
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.